LLM-based NTU course recommendation systems

Abstract

One of the most pivotal decisions that higher education students frequently make is the planning and selection of their courses at the start of every academic semester. Course selection directly influences their ability to meet degree requirements and significantly impact their academic performance. Furthermore, with the extensive range of courses provided, students are often overwhelmed by the abundance of information and may face analysis paralysis. Hence, there is an increasing demand for innovative solutions like Course Recommendation Systems (CRS), which leverage technology to offer personalised, data-driven guidance.

The primary aim of this project is to design an intelligent chatbot that will answer questions related to course planning and empower students to be more strategic in charting their academic trajectories.

This project utilises Large Language Models (LLMs) with the latest web and cloud technologies, including Streamlit for Frontend, LangChain and LangGraph to connect with LLMs application, CosmosDB for a serverless database, Neo4j to build knowledge graph, as well as Azure AI Search for the implementation of VectorRAG. Such an approach aims to streamline the setup of the intelligent chatbot and revolutionise the way users interact with digital content. This system will serve as a comprehensive, one-stop service for answering student queries, ultimately becoming their personalised academic companion.